Monday, 18 February 2013

Converting an ATX power supply for bench use

Somewhere amidst my haze of academia and cheap beer, I managed to cook my old bench PSU (also ATX).  Unfortunately I built the thing worse than a shit brickhouse, and it shorted out about a hundred times before it decided to stop living.  I'm sorry, old pal.

Being an electronics enthusiast, not having a bench supply was a total pain in the ass.  Luckily, old ATX power supplies are about as prevalent amongst computer geeks as chlamydia is amongst arts students -  so, me and a few fellow nerds gathered some up, ordered a few parts off eBay, and made fast work of building some chlamydia machines power supplies!

Fig. 1 The victim - cover removed

The first step is to lop off the wiring harness, leaving enough slack in the wires to solder onto the binding posts later on (Fig 2).  I like to save the leftover wires to turn into test leads (alligator clips, etc).

Fig. 2 The victim - castrated
Next, you need to find a good spot to place your binding posts.  This is crucial, and can be difficult, depending on the model.  You want to be careful that your binding posts aren't in contact with any heatsinks, capacitors, etc, when you place the cover back on.  Also, you want to make sure the remaining lengths of wire are able to reach them - and then some.  You want to leave enough slack for when you manoeuvre the cover back on.

Also worth noting, you need to get binding posts that have insulated mounting hardware.  Otherwise, the electrical contacts are liable to wiggle loose and short your supply to the grounded casing.  This is the fate that befell my old supply, as I was dense enough to use cheap non-insulated binding posts... derp.

Fig. 3 Eyeballed where to place the posts

Making holes for the binding posts is easy enough, use a hammer and punch screw to define your holes for best results.

Fig. 4 "This isn't exactly an exact Science" - Matt K, B.Sc.E.E., hammering a screw

Fig. 5 - Not exactly uniform, whatever.  Wasn't pretty in the first place.

Once we dry-fitted the binding posts, we realized that the mounting gap in our binding posts was larger than the thickness of the power supply casing.  And if there's one thing you don't want in your chlamydia machine, it's loose plugs jiggling all up in your holes.  Matt M fixed our size issues by cutting chunks of plastic hose to use as spacers - Ghetto, but effective!

Fig. 6 - Making spacers out of plastic tubing
Somehow, I failed to realize that my supply didn't have an on/off switch on the back, like most PSUs.  I fixed this by putting a switch between the power-on signal and a ground wire (green and black wires from the main mobo connector, respectively).  If the PSU had a switch already, I would have just shorted the power-on and ground wires.  

Fig. 7 - All mounted up, including the new switch

It is advisable that you always have a load present on these ATX supplies, as it is a switch-mode supply.  Not having a load causes it to go into discontinuous conduction mode, or something like that - I don't recall the exact details, but just slap a high-wattage, low-resistance resistor between a 3.3V wire and ground and you'll be good. Here we used a 12 ohm.  You want to ensure that your selected resistance is within the bounds of your supply; 3.3V @ 12 ohms is roughly 1/4 A, and the 3.3V supply is rated for 20A, so we are well within our limits.

For good measure, I insulated my load resistor's leads with some heat shrink tubing.

Fig. 8 Constant load resistor, heat shrink sleeve for isolation

Fig. 9 - Resistor heat shrinked, ready to be carelessly crammed into place

It is also advisable to cap off all the extra connections that went unused. Our approach was to use heat-shrink tubing to 'group' together the different voltage supply lines (for example, all ground wires grouped together, all 5V wires grouped together)

Fig. 10 Terminals taped up for good measure, extra wires capped off with heat shrink

Now all that's left is to slap it back together and label the thing.  Apparently I really suck at this (Fig 11).

Fig. 11 Wasn't exactly on my A-game...

For best results, you need a really badass name for the thing.  Pro tip: anything that ends in "5000" is totally badass.

Fig. 12 Its all in the name.
For the hell of it, we decided to look at the ripple on each supply line.  Generally speaking, all of the supply voltages had an unloaded ripple of < 10 mV - not bad at all!  The +/- 12V lines had a little more, around 20mV for my PSU.

Fig. 13 Matt K checking the ripple

Also, nothing helps your fabrication skills more than a cat to move your shit around and stare at you with cold, evil intentions.  This keeps you on your toes, able to react to any unforeseen fuck-ups with greater acuity.  Praise thee, feline superbeings!

Fig. 14 Matt M's cat treat launcher is rumoured to also work as a solder sucker

No comments:

Post a Comment